feat: device hot-reload, tools-relative adb, auto port bind #3586
Conversation
packages/mobile/src/lib.rs
Outdated
| } | ||
| } | ||
|
|
||
| static APP_FN_PTR: Mutex<Option<fn() -> Element>> = Mutex::new(None); | ||
| static APP_OBJECTS: Mutex<Option<BoundLaunchObjects>> = Mutex::new(None); |
There was a problem hiding this comment.
Could this be a thread local instead of a static? Box<dyn Any> is not send or sync so this looks unsound
There was a problem hiding this comment.
I need to test on-device to see if the android on_load is guaranteed to be fired from the main thread...
there's a chance it's not and we might need to mess with how this is structured a bit.
There was a problem hiding this comment.
Okay the way we do main has weird semantics. I think technically this is fine since the program needs to terminate before the main() call can finish up during load. I believe this the same case as an Rc moving between threads being sound only if you can prove it moves completely. We actually do have an assurance the Box<dyn Any> is not held onto anything from the main thread since the main thread no longer exists.
I'll add a note about this in the code.
Fixes a number of issues:
dx serve) #3444dx servedoes not build if the process is backgrounded #3522 backgrounded issueAttempted and failed to get code-signing working. My old prototype doesn't work anymore.
This might get us close to polishing the number of CLI bugs I'm currently aware of.